🚗 CAV Security Game Theory Simulations

Interactive simulations for analyzing security decisions in Connected and Autonomous Vehicles (CAVs) using game-theoretic models.

This project is a Streamlit-based application that models cybersecurity scenarios in automotive networks. It provides interactive visualizations and solvers for two specific game theory models: Stackelberg Security Games for IDS placement and Bayesian Signaling Games for GPS spoofing detection.

🌟 Key Features

1. 🛡️ IDS Placement Game (Stackelberg Game)

Models the interaction between a defender placing Intrusion Detection Systems (IDS) and an attacker targeting Electronic Control Units (ECUs).

• Game Type: Leader-Follower (Stackelberg). The defender commits to a defense strategy, and the attacker observes and optimizes their attack.
• Simulation Features:
  • Network Topology: Visualizes the vehicle network and protected nodes.
  • Algorithms: Compare three different placement strategies:
    • Exact: Exhaustive search (optimal for small N).
    • Greedy Heuristic: Fast, near-optimal placement based on criticality.
    • Genetic Algorithm: Evolutionary approach for complex spaces.
  • Analysis: Comparison of computation time, defender payoff, and attack targets.
  • Defense Performance Comparison: Damage reduction metrics vs no-defense baseline with algorithm performance summary.

2. 📡 GPS Spoofing Game (Bayesian Game)

Models a scenario where an attacker may inject spoofed GPS signals, and the defender must decide whether to verify the signal based on observed deviations.

• Game Type: Bayesian Signaling Game with incomplete information. The defender uses Bayes' rule to update their belief about whether a signal is benign noise or a malicious attack.
• Simulation Features:
  • Belief Updates: Visualizes how the probability of an attack changes based on signal deviation.
  • Equilibrium Analysis: Determines if the game state results in a Separating, Pooling, or Semi-Separating equilibrium.
  • ROC Analysis: Receiver Operating Characteristic curves for detection thresholds.
  • Repeated Games: Simulates the game over multiple rounds to track cumulative payoffs and detection rates.

📂 Project Structure

.
├── app.py                      # Main entry point for the Streamlit application
├── requirements.txt            # Python dependencies
├── algorithms/                 # Game theory logic and solvers
│   ├── __init__.py
│   ├── stackelberg.py          # Solvers for IDS placement (Exact, Greedy, Genetic)
│   ├── bayesian.py             # Logic for belief updates and Bayesian equilibrium
│   └── utils.py                # Helper functions (payoffs, plotting, etc.)
└── pages/                      # Streamlit multipage files
    ├── 1_IDS_Placement_Game.py # UI for the IDS Simulation
    └── 2_GPS_Spoofing_Game.py  # UI for the GPS Spoofing Simulation

🚀 Installation & Usage

Prerequisites

• Python 3.8 or higher

1. Clone the Repository

git clone <repository-url>
cd GameTheory-Simulations

2. Create a Virtual Environment (Optional but Recommended)

# Windows
python -m venv venv
.\venv\Scripts\activate

# macOS/Linux
python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Run the Application

streamlit run app.py

The application will launch in your default web browser (typically at http://localhost:8501).

📚 Theoretical Background

Stackelberg Equilibrium

In the IDS Placement Game, the Defender acts as the Leader, committing to a randomized allocation of IDS resources. The Attacker acts as the Follower, observing the defense strategy and attacking the ECU that maximizes their utility. The solver finds the allocation that maximizes the Defender's utility assuming the Attacker plays optimally.

Bayesian Belief Update

In the GPS Spoofing Game, the Defender observes a signal deviation $d$. They update their belief $\mu$ (probability the signal is malicious) using Bayes' Rule:

$$\mu(\text{Malicious} | d) = \frac{P(d | \text{Malicious}) \times P(\text{Malicious})}{P(d)}$$

The Defender verifies the signal only if the posterior belief exceeds a calculated threshold $\tau$.

🛠️ Built With

• Streamlit: Interactive web interface.
• Plotly: Interactive charts and network graphs.
• NetworkX: Graph topology for ECU networks.
• NumPy & SciPy: Mathematical computations and statistical distributions.